x86/string: Fixup alignment of main loop in str{n}cmp-evex [BZ #32212]

The loop should be aligned to 32-bytes so that it can ideally run out
the DSB. This is particularly important on Skylake-Server where
deficiencies in it's DSB implementation make it prone to not being
able to run loops out of the DSB.

For example running strcmp-evex on 200Mb string:

32-byte aligned loop:
    - 43,399,578,766      idq.dsb_uops
not 32-byte aligned loop:
    - 6,060,139,704       idq.dsb_uops

This results in a 25% performance degradation for the non-aligned
version.

The fix is to just ensure the code layout is such that the loop is
aligned. (Which was previously the case but was accidentally dropped
in 84e7c46df).

NB: The fix was actually 64-byte alignment. This is because 64-byte
alignment generally produces more stable performance than 32-byte
aligned code (cache line crosses can affect perf), so if we are going
past 16-byte alignmnent, might as well go to 64. 64-byte alignment
also matches most other functions we over-align, so it creates a
common point of optimization.

Times are reported as ratio of Time_With_Patch /
Time_Without_Patch. Lower is better.

The values being reported is the geometric mean of the ratio across
all tests in bench-strcmp and bench-strncmp.

Note this patch is only attempting to improve the Skylake-Server
strcmp for long strings. The rest of the numbers are only to test for
regressions.

Tigerlake Results Strings <= 512:
    strcmp : 1.026
    strncmp: 0.949

Tigerlake Results Strings > 512:
    strcmp : 0.994
    strncmp: 0.998

Skylake-Server Results Strings <= 512:
    strcmp : 0.945
    strncmp: 0.943

Skylake-Server Results Strings > 512:
    strcmp : 0.778
    strncmp: 1.000

The 2.6% regression on TGL-strcmp is due to slowdowns caused by
changes in alignment of code handling small sizes (most on the
page-cross logic). These should be safe to ignore because 1) We
previously only 16-byte aligned the function so this behavior is not
new and was essentially up to chance before this patch and 2) this
type of alignment related regression on small sizes really only comes
up in tight micro-benchmark loops and is unlikely to have any affect
on realworld performance.

Reviewed-by: H.J. Lu <hjl.tools@gmail.com>

sysdeps/x86_64/multiarch/strcmp-evex.S

@@ -209,7 +209,9 @@
    returned.  */
 
 	.section SECTION(.text), "ax", @progbits
-	.align	16
+	/* Align 64 bytes here. This is to get the L(loop) block ideally
+	   aligned for the DSB.  */
+	.align	64
 	.type	STRCMP, @function
 	.globl	STRCMP
 # ifdef USE_AS_STRCASECMP_L
@@ -509,9 +511,7 @@ L(ret4):
 	ret
 # endif
 
-	/* 32 byte align here ensures the main loop is ideally aligned
-	   for DSB.  */
-	.p2align 5
+	.p2align 4,, 4
 L(more_3x_vec):
 	/* Safe to compare 4x vectors.  */
 	VMOVU	(VEC_SIZE)(%rdi), %VMM(0)
@@ -1426,10 +1426,9 @@ L(less_32_till_page):
 L(ret_zero_page_cross_slow_case0):
 	xorl	%eax, %eax
 	ret
-# endif
-
-
+# else
 	.p2align 4,, 10
+# endif
 L(less_16_till_page):
 	cmpl	$((VEC_SIZE - 8) / SIZE_OF_CHAR), %eax
 	ja	L(less_8_till_page)
@@ -1482,8 +1481,12 @@ L(less_16_till_page):
 # endif
 	jmp	L(prepare_loop_aligned)
 
-
-
+# ifndef USE_AS_STRNCMP
+	/* Fits in aligning bytes.  */
+L(ret_zero_4_loop):
+	xorl	%eax, %eax
+	ret
+# endif
 
 	.p2align 4,, 10
 L(less_8_till_page):
@@ -1554,6 +1557,7 @@ L(ret_less_8_wcs):
 
 #  ifdef USE_AS_STRNCMP
 	.p2align 4,, 2
+L(ret_zero_4_loop):
 L(ret_zero_page_cross_slow_case1):
 	xorl	%eax, %eax
 	ret
@@ -1586,10 +1590,6 @@ L(less_4_loop):
 	subq	$-(CHAR_PER_VEC * 4), %rdx
 #  endif
 	jmp	L(prepare_loop_aligned)
-
-L(ret_zero_4_loop):
-	xorl	%eax, %eax
-	ret
 L(ret_less_4_loop):
 	xorl	%r8d, %eax
 	subl	%r8d, %eax